Human recognition method based on privacy protection

Abstract

The human body recognition method based on privacy protection solves the problem of how to protect privacy in monitoring video, and belongs to the field of electronic information engineering and image processing.The present application comprises: S1, using a human body recognition model to recognize the human body in the range of the camera, and drawing a rectangular frame for each human body appearing in the camera; S2, performing multiple blur processing on the human body image in the rectangular frame, and performing fuzzy display on the human body appearing in the camera.The human body in the rectangular frame is subjected to multiple Gaussian blur processing, and the blurred human body cannot be directly recognized.This privacy protection function can effectively avoid the leakage of user privacy and protect the legitimate rights and interests of the user.

Description

Technical Field

[0001] This invention relates to a privacy-preserving human body recognition method, belonging to the fields of electronic information engineering and image processing. Background Technology

[0002] Human body recognition technology is increasingly being applied in various fields, such as security, transportation, and healthcare. It not only has broad application prospects but also plays an increasingly important role in our daily lives. Essentially, this technology protects user privacy during video surveillance. With continuous technological advancements, human body recognition technology has been continuously improved and applied. Among these, deep learning-based human detection algorithms are an important type of target detection algorithm within human body recognition. Although traditional object detection algorithms still require a large amount of manually labeled data and computational resources, it is undeniable that deep learning-based human detection algorithms have become a research hotspot in the current field of target detection.

[0003] Many homes now have live webcams installed, and the images recorded often involve significant private information. Therefore, how to handle these private videos and protect users' privacy rights is a crucial issue. If criminals obtain these private video streams, it can cause irreparable damage to users' rights. Summary of the Invention

[0004] To address the issue of protecting privacy in surveillance videos, this invention provides a privacy-preserving human body recognition method.

[0005] The present invention provides a privacy-preserving human body recognition method, comprising:

[0006] S1. Use a human body recognition model to identify human bodies within the camera's range and draw a rectangular frame for each human body appearing in the camera's view.

[0007] S2. Perform multiple blurring processes on the human image within the rectangular frame to blur the human body appearing in the camera's view.

[0008] Preferably, in step S2, the human image within the rectangular frame is blurred, including:

[0009] The human image within the rectangle is used as the original image. The original image is then subjected to Laplacian blurring. Finally, a sparse coding algorithm is used to sparsely represent the blurred image to obtain a sparse image.

[0010] The sparse image is reverse-processed to generate a new blurred image. Then, a sparse coding algorithm is used to sparsely represent the new blurred image to obtain the sparse image. This process is repeated until a set number of iterations is reached, at which point the iteration stops.

[0011] After a set number of iterations, the sparse image obtained is fused with the original image. The resulting blurred fused image is input into the generator. The blurred image output by the generator and the original image are simultaneously input into the discriminator. The discriminator outputs the blurred image, thus completing the blurring process.

[0012] The generator and discriminator form a generative adversarial network.

[0013] Preferably, the human body recognition model is based on the YOLOv3 network.

[0014] As a preferred method, the frame difference method is used to identify moving objects within the camera's range, and then Kalman filtering is used for tracking. The tracked target is then input into the human body recognition model.

[0015] Preferably, the human body image is a face image, and face detection is performed using Haar cascade classification.

[0016] The beneficial effects of this invention are that it applies multiple Gaussian blurring processes to the human body within the rectangular frame, making the body impossible to directly identify. This privacy protection function effectively prevents the leakage of user privacy and protects users' legitimate rights and interests. For example, in public places such as shopping malls and schools, this system can quickly detect and handle violations while preventing the leakage of personal privacy. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the principle of the present invention;

[0018] Figure 2 This is a blurred effect diagram of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0022] The privacy-preserving human identification method of this embodiment includes:

[0023] Step 1: Use a human body recognition model to identify people within the camera's range and draw a bounding box for each person appearing in the camera's view. This step involves designing a human body recognition model that protects user privacy based on a deep learning network. By training on the COCO dataset, an efficient and accurate model is obtained. Then, this model is used to identify people within the camera's range and draw a bounding box for each person appearing in the camera's view.

[0024] S2. The human image within the rectangular frame was blurred multiple times, and the human body appearing in the camera's view was displayed in a blurred manner.

[0025] This step involves multiple blurring processes applied to the human body within the rectangular frame, making it impossible to directly identify. This privacy protection feature effectively prevents the leakage of user privacy and protects users' legitimate rights. For example, in public places such as shopping malls and schools, this system can quickly detect and handle violations while preventing the leakage of personal privacy.

[0026] In a preferred embodiment, step 2 involves multiple blurring processes on the human image within the rectangular frame, including:

[0027] The human image within the rectangle is used as the original image. The original image is then subjected to Laplacian blurring. Finally, a sparse coding algorithm is used to sparsely represent the blurred image to obtain a sparse image.

[0028] The sparse image is reverse-processed to generate a new blurred image. Then, a sparse coding algorithm is used to sparsely represent the new blurred image to obtain the sparse image. This process is repeated until a set number of iterations is reached, at which point the iteration stops.

[0029] After a set number of iterations, the sparse image obtained is fused with the original image. The resulting blurred fused image is input into the generator. The blurred image output by the generator and the original image are simultaneously input into the discriminator. The discriminator outputs the blurred image, thus completing the blurring process.

[0030] The generator and discriminator form a generative adversarial network.

[0031] To protect user privacy, the human body within the bounding box is blurred using a novel adaptive algorithm. This algorithm first alternately applies sparse coding and Laplacian blurring to obtain a better blurred image. Specifically, it first blurs the input image, then uses sparse coding to sparsely represent the blurred image. Next, it uses a blurring algorithm to reverse-process the sparse signal and generate a new blurred image. This process can be iterated multiple times to obtain better results. Simultaneously, a generative adversarial network is integrated, employing a conditional generative adversarial network (cGAN): in this method, the blurred image obtained from sparse coding and the Laplacian algorithm is used as input to the generator, and the generator's output is a more blurred image. This blurred image, along with the original image, is then used as input to the discriminator. By continuously adjusting the generator and discriminator, the generator can generate increasingly blurred images. The modulated model is then applied to the process of blurring the human body, maximizing the stability and high accuracy of the blurring effect.

[0032] In step 1 of this implementation method, a human body recognition model is established based on the YOLOv3 network.

[0033] In a preferred embodiment, the frame difference method is used to identify moving objects within the camera's range, and then Kalman filtering is used for tracking. The tracked target is then input into the human body recognition model.

[0034] The biggest problem in the field of human detection and even object detection in the industry is the limitation of detection accuracy. Most systems suffer from detection distortion and errors in a single frame of the video stream, which poses a risk of exposing human bodies whose privacy needs to be protected. By organically combining the frame difference method with Kalman filtering and integrating it with the YOLO network model, we can achieve the detection and localization of moving targets within the video range, thereby blurring all moving objects in the video stream. This provides a good auxiliary blurring effect for moving human bodies in the video.

[0035] In this embodiment, the human body image is a face image, and Haar cascade classification is used for face detection. Since the face is the most private part of the human body, this embodiment also incorporates a face blurring effect, using a lightweight convolutional neural network architecture and optimized algorithms to achieve a good auxiliary blurring effect while ensuring the smoothness of the video stream and the speed of computation.

[0036] The aforementioned privacy-preserving human body recognition method is developed into a computer program. The entire program is packaged into a separate folder; clicking the .exe file within the folder allows the program to be run locally, achieving portability and universal applicability. A GUI interface module is also designed. Clicking the .exe file pops up a window for entering a key. Only by entering the correct key can access the monitoring system and view the image generated by the privacy-preserving human body recognition method. Entering an incorrect key or closing the password input window will prevent access to the system, providing a simple encryption operation. Testing has shown that the system is efficient, accurate, and privacy-preserving, suitable for various scenarios requiring human body recognition and privacy protection.

[0037] This implementation packages the entire file into a single folder. Clicking the .exe file within the folder allows the program to be run locally, achieving portability and universal applicability. This allows the system to run independently on various devices, making it more convenient for users. For example, users can run the system on their own computers or mobile devices without installing additional software or hardware.

[0038] Application Example 1: Applying this application in private residences can effectively protect residents' privacy and improve home security. The following are some specific implementation methods:

[0039] Installing security cameras: First, you need to install security cameras within your home. The number and location of the cameras should be determined based on your specific needs. Generally, they should be installed in easily accessible areas such as the entrance, hallway, and living room. Additionally, you should choose high-definition cameras for better environmental detection.

[0040] Automatic loading system: After installing the cameras, this application needs to be set up. Simply download the inventor's relevant folder and copy it to the computer where each camera is located, then set it to start automatically on boot. This ensures the system is always in monitoring mode.

[0041] Backup and Updates: To ensure system stability and security, system data should be backed up regularly, and the system should be updated and upgraded as needed to ensure that it has the latest functions and security.

[0042] Through the above implementation plan, this application can be successfully applied in private residences to protect residents' privacy and family security.

[0043] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A human body recognition method based on privacy protection, characterized by, The method includes: S1. Use a human body recognition model to identify human bodies within the camera's range and draw a rectangular frame for each human body appearing in the camera's view. S2. Perform multiple blurring processes on the human image within the rectangular frame to blur the human body appearing in the camera's view. In S2, the human image within the rectangular frame is blurred, including: The human image within the rectangle is used as the original image. The original image is then subjected to Laplacian blurring. Finally, a sparse coding algorithm is used to sparsely represent the blurred image to obtain a sparse image. The sparse image is reverse-processed to generate a new blurred image. Then, a sparse coding algorithm is used to sparsely represent the new blurred image to obtain the sparse image. This process is repeated until a set number of iterations is reached, at which point the iteration stops. After a set number of iterations, the sparse image obtained is fused with the original image. The resulting blurred fused image is input into the generator. The blurred image output by the generator and the original image are simultaneously input into the discriminator. The discriminator outputs the blurred image, thus completing the blurring process. The generator and discriminator form a generative adversarial network.

2. The privacy-preserving human identification method of claim 1, wherein, The human body recognition model is established based on the YOLOv3 network in S1. 3.The privacy-preserving human identification method of claim 1, wherein, The frame difference method is used to identify moving objects within the camera's range, and then Kalman filtering is used for tracking. The tracked targets are then input into the human body recognition model. 4.The privacy-preserving human identification method of claim 1, wherein, The human body image is a face image, and Haar cascade classification is used for face detection.

5. A storage device readable by a computer, the storage device storing a computer program, characterized in that, When the computer program is executed, it implements the privacy-preserving human identification method as described in any one of claims 1 to 4.

6. A privacy-preserving human identification system comprising a storage device, a processor, and a computer program stored in the storage device and executable on the processor, wherein, The processor executes the computer program to implement the privacy-preserving human identification method as described in any one of claims 1 to 4.

7. The privacy-preserving human body recognition system according to claim 6, characterized in that, The system also includes a GUI interface module, which is used to input a key. When the key is correct, the camera is turned on and the image is displayed after the processor executes the computer program to implement the privacy-preserving human body recognition method as described in any one of claims 1 to 4.

Images